Support NeoGAF

[benny_a]

I got demos of most of these at E3. I saw Watch_Dogs in person three times — and the demo went differently each time, by the way, and they covered different ground for the same objective at the end, so it wasn't especially scripted — and it's much more impressive in person than any video of it online. Video compression kicks the shit out of the IQ.

Sorry for being off-topic but was that embargoed? I don't remember hearing about that part.
The first thought many people had when seeing Watch Dogs is: Looks fantastic, hopefully it isn't very scripted.
And now you're saying it isn't is big news to me.

Pop-in and LOD transitions go hand-in-hand here, two symptoms of failed attempts to solve the same optimization problem. Tesselation should hopefully mitigate that, if not remove it entirely.

Could you expand or point me to the direction of how Tesselation can help with LOD transitions and pop-in?

Showing my ignorance, but all I've ever seen was bricks (and other things with those convex characteristics) looking better with it turned on.

 

[Persona86]

I'm not a graphics whore, but this thread is getting me excite.

 

[shinobi602]

Sorry for being off-topic but was that embargoed? I don't remember hearing about that part.
The first thought many people had when seeing Watch Dogs is: Looks fantastic, hopefully it isn't very scripted.
And now you're saying it isn't is big news to me..

Ubisoft said that a long time ago.

 

[Zyrusticae]

Could you expand or point me to the direction of how Tesselation can help with LOD transitions and pop-in?

Showing my ignorance, but all I've ever seen was bricks (and other things with those convex characteristics) looking better with it turned on.

Tesselation can be used as a replacement for the LOD systems of the past. By controlling the triangle count for objects at certain distances (with objects that only occupy a few pixels on the screen being reduced to a few hundred polygons or less), you can get great performance without suffering from all the pitfalls of traditional LOD systems, including the actual generation of said LOD assets.

This, in my mind, is the biggest boon of tesselation - more than just increasing the detail of assets up-close, it can completely change the way we optimize our games.

Best of all, it's very flexible - you can change the triangle count dynamically, which means it's quite scalable and can be made to run well on a wide range of systems.

Edit: Here, look at Nvidia's "endless city" demonstration for an idea of what it can be used for.

 

[rottendevice]

If I recall correctly I think Carmack said the basis would be fully customizable geometry. Basically taking the idea of mega textures and applying it to building geometry. I could be wrong on that though.

Oh neat. I guess I meant, we haven't seen any screenshots yet.

 

[RoboPlato]

Tesselation can be used as a replacement for the LOD systems of the past. By controlling the triangle count for objects at certain distances (with objects that only occupy a few pixels on the screen being reduced to a few hundred polygons or less), you can get great performance without suffering from all the pitfalls of traditional LOD systems, including the actual generation of said LOD assets.

This, in my mind, is the biggest boon of tesselation - more than just increasing the detail of assets up-close, it can completely change the way we optimize our games.

Best of all, it's very flexible - you can change the triangle count dynamically, which means it's quite scalable and can be made to run well on a wide range of systems.

Edit: Here, look at Nvidia's "endless city" demonstration for an idea of what it can be used for.

Yeah, I'm really excited to see what devs do with tesselation next gen. With it being readily available in consoles I think we'll finally start to see it used for more than just making bricks and silhouette refinement (although we'll see a lot of that). LoD use, like you mentioned, is incredibly promising and its use in water simulation could be fantastic.

 

[i-Lo]

Tesselation can be used as a replacement for the LOD systems of the past. By controlling the triangle count for objects at certain distances (with objects that only occupy a few pixels on the screen being reduced to a few hundred polygons or less), you can get great performance without suffering from all the pitfalls of traditional LOD systems, including the actual generation of said LOD assets.

This, in my mind, is the biggest boon of tesselation - more than just increasing the detail of assets up-close, it can completely change the way we optimize our games.

Best of all, it's very flexible - you can change the triangle count dynamically, which means it's quite scalable and can be made to run well on a wide range of systems.

Edit: Here, look at Nvidia's "endless city" demonstration for an idea of what it can be used for.

I am curious, are current gen consoles in any way capable of tessellation?

 

[Thunder Monkey]

I am curious, are current gen consoles in any way capable of tessellation?

Wind Waker used it for the waves on the GCN.

Tessellation is just a catch all term generally. Seeing it used as it was intended though? Not really. Turns out the answer was "More tessellators!"

edit: Yes I'm kind of being a smart ass.

 

[JordanN]

I am curious, are current gen consoles in any way capable of tessellation?

They can but it's quite limited compared to the Directx 11 stuff.

For example, Viva Pinata uses it for the soil.

 

[i-Lo]

Wind Waker used it for the waves on the GCN.

Tessellation is just a catch all term generally. Seeing it used as it was intended though? Not really. Turns out the answer was "More tessellators!"

edit: Yes I'm kind of being a smart ass.

They can but it's quite limited compared to the Directx 11 stuff.

For example, Viva Pinata uses it for the soil.

Ah thanks.

And Thunder Monkey, your posts are a pleasure to read, as always.

 

[SniperHunter]

I can't wait for full tessellation support, lod is such a pos. It's power hungry and eats a lot of FPS, tesellation on the other hand seems great. Agent 47 tessellated bald head looks sooo good!

 

[kevinski]

I'm unable to start new threads, so I figured that this question might fit into a thread about next-gen game engines. Basically, a co-worker of mine who happens to be really into video games was telling me about his dream game of sorts. He has no interest in making the game himself, so I was questioning the viability of a game like this. He'd like to see a next-gen surfing game with really accurate water physics. Do any existing or future engines seem to be putting much emphasis on accurate water physics to the point that they could power an ultra-realistic surfing game, preferably in an open-world setting?

 

[Chev]

I'm unable to start new threads, so I figured that this question might fit into a thread about next-gen game engines. Basically, a co-worker of mine who happens to be really into video games was telling me about his dream game of sorts. He has no interest in making the game himself, so I was questioning the viability of a game like this. He'd like to see a next-gen surfing game with really accurate water physics. Do any existing or future engines seem to be putting much emphasis on accurate water physics to the point that they could power an ultra-realistic surfing game, preferably in an open-world setting?

The tech exists, this GDC talk explains it all, basically spawned many game uses like From Dust. Most game implementations only use non-breaking waves because that's much simpler though, and works pretty well in many case. Specifically existing and close to what you're searching for is Sail Simulator.

There there's some steps up from that same method: this one, available in Physx, integrates some breaking, but this one's the absolute best for future game use so far (go to 2:40 for the crazy waves). But note I say future, I'm not aware of any actual use or how well it'd fare with a game around it with current tech. But it's coming, definitely!

 

[kevinski]

The tech exists, this GDC talk explains it all, basically spawned many game uses like From Dust. Most game implementations only use non-breaking waves because that's much simpler though, and works pretty well in many case. Specifically existing and close to what you're searching for is Sail Simulator.

There there's some steps up from that same method: this one, available in Physx, integrates some breaking, but this one's the absolute best for future game use so far (go to 2:40 for the crazy waves). But note I say future, I'm not aware of any actual use or how well it'd fare with a game around it with current tech. But it's coming, definitely!

Thanks for the info! Nice to see that there's interest in something like this.

 

[childplease]

PC Gaming has truly spoiled me.

None of those next gen engine pics impressed me :/

Cmon next next gen come out already.

 

[mrklaw]

I can't wait for full tessellation support, lod is such a pos. It's power hungry and eats a lot of FPS, tesellation on the other hand seems great. Agent 47 tessellated bald head looks sooo good!

isn't tessellation quite power hungry too? Is it a balance between more memory needed for multiple LoD models, bandwidth for switching out various models vs GPU grunt for tessellating

 

[Eideka]

PC Gaming has truly spoiled me.

None of those next gen engine pics impressed me :/

And yet they are miles ahead of everything available right now.

 

[Backfoggen]

PC Gaming has truly spoiled me.

None of those next gen engine pics impressed me :/

Cmon next next gen come out already.

I wanna see the PC games you've played.

 

[Reiko]

I wanna see the PC games you've played.

He's on Direct X20.

 

[Glass Rebel]

To be honest, I find the presentation slides to most of these engines even more impressive although I only understand a fraction of it.

 

[pottuvoi]

Funny thing is IIRC this lighting was cut from BF3, even on the PC. Not sure if it's been used in any of the other FB2 games.

I'm pretty sure it is still used for everything else than dynamic lights.

 

[shinobi602]

PC Gaming has truly spoiled me.

None of those next gen engine pics impressed me :/

Cmon next next gen come out already.

Still waiting for those PC games you've been playing.

 

[Zarx]

Was there any indication if Just Cause 3 (or whatever Avalanche Studios is working on now) will use proprietary engine?

Well they have always used their preparatory Avalanche engine before, I don't see why that would change. They have 2 studios now and they are both working on next gen games.

I can't wait to see what Avalanche 3.0 can do.

 

[Sid]

Still waiting for those PC games you've been playing.

I think he was being sarcastic.

 

[FACE]

What about Star Citizen(CryEngine 3)?

Edit: More screens and video.

 

[Zarx]

What about Star Citizen(CryEngine 3)?

Star Citizen looks stunning for sure, tho I hope you can turn the motion blur off.

 

[i-Lo]

What about Star Citizen(CryEngine 3)?

Honestly, I was in awe when I saw the game. The only thing detracting from atm (primarily because it's incomplete) are the animations of people.

Amazing results achieved with Cryengine 3

 

[Zyrusticae]

isn't tessellation quite power hungry too? Is it a balance between more memory needed for multiple LoD models, bandwidth for switching out various models vs GPU grunt for tessellating

He got it wrong, tessellation in no way saves power (though it can decrease strain on a card if developers choose to allow tessellation towards low polygon counts) relative to LOD systems.

The major benefit is that it looks a thousand times better than current LOD systems. LOD systems of today are the reason games have so much bloody pop-in problems - well, that, and inability to properly instance foliage such that the frame rate doesn't tank when there's hundreds of thousands of blades of grass on the screen.

Tessellation solves the former problem. For the latter, we'll see...

 

[Zarx]

He got it wrong, tessellation in no way saves power (though it can decrease strain on a card if developers choose to allow tessellation towards low polygon counts) relative to LOD systems.

The major benefit is that it looks a thousand times better than current LOD systems. LOD systems of today are the reason games have so much bloody pop-in problems - well, that, and inability to properly instance foliage such that the frame rate doesn't tank when there's hundreds of thousands of blades of grass on the screen.

Tessellation solves the former problem. For the latter, we'll see...

Crysis 3 has a lot of grass lol, so maybe they have solved that issue to an extent but I guess we will have to wait and see how performance holds up.

Edit: Just remembered Bitsquid
Scalable rendering pipe with multiple back-ends.
Deferred shading, physical lighting, subsurface scattering, hardware tessellation, DOF, SSAO, tone-mapping, 3D vision, etc.
All bulk workloads in the engine are run through a job system which provides data parallelism for an arbitrary number of CPU cores or co-processors (SPUs).
Short iteration times. No baking.
Drop-in play on PC and consoles from level editor.

 

[Nealand Liquor]

I know I'm running the risk of sounding like a broken record in threads like these, but do you think fully destructable environments will be possible next gen on console or PC?
Now I don't mean destructable as in an item here, a wall there, ala BF3/Force Unleashed etc; But fully destructable everything, as in every single thing can be manipulated, destorted and destroyed - or at the very least the great lions share?

As a follow up: if not what are our current limitations?

 

[LowMax]

I know I'm running the risk of sounding like a broken record in threads like these, but do you think fully destructable environments will be possible next gen on console or PC?
Now I don't mean destructable as in an item here, a wall there, ala BF3/Force Unleashed etc; But fully destructable everything, as in every single thing can be manipulated, destorted and destroyed - or at the very least the great lions share?

As a follow up: if not what are our current limitations?

We need to be able to simulate matter itself if you want everything to be able to be manipulated. Hello holodeck.

 

[Zaptruder]

I know I'm running the risk of sounding like a broken record in threads like these, but do you think fully destructable environments will be possible next gen on console or PC?
Now I don't mean destructable as in an item here, a wall there, ala BF3/Force Unleashed etc; But fully destructable everything, as in every single thing can be manipulated, destorted and destroyed - or at the very least the great lions share?

As a follow up: if not what are our current limitations?

Information modelling.

Essentially, we'd have to build detailed internal models for things to not look dumb when you destroy them. Too much stuff that you can't guarantee will be seen at all.

e.g. blow open a computer; you'd expect computer parts in it. Blow open a fridge, you'd expect food in there. Blow open the fridge and splatter the food... you'd expect the food to splatter realistically against the wall, with bits of food scattered about.

Blow open a glass and steel door - you'd expect shattered glass and twisted door frame.

Until we design a system that allows us to do information modelling extremely quickly (i.e. we build a door as a door actually would be built, we build a building with the internal structure as it would be built, we give them all their expected material properties (i.e. cloth burns and shreds like cloth, metal twists and melts like metal)...

It's just going to be easier to limit destruction.

Given what I'm describing... I'd say that the kind of tech that would allow that to be market feasible and tech feasible at real time speeds is around 10 to 15+ years away.

Of course, we'll get more procedural information details along the way, so that by the time it does happen... we'll have spent the last 5 years before that able to blow up a lot of things, but not everything.

 

[Nealand Liquor]

Hmmmm... I see your point, but isn't there middle ground we can reach next gen?
In the short term (5-10 yrs) I'd be happy if we kept the current way we handle the visual of destruction, but applied it to absolutely everything.

 

[Zaptruder]

Hmmmm... I see your point, but isn't there middle ground we can reach next gen?
In the short term (5-10 yrs) I'd be happy if we kept the current way we handle the visual of destruction, but applied it to absolutely everything.

Well... you don't see my point, because my point is the current level of visual destruction is hand modelled, and it's too expensive to hand model that for everything.

 

[shinobi602]

Well... you don't see my point, because my point is the current level of visual destruction is hand modelled, and it's too expensive to hand model that for everything.

Even in Frostbite?

 

[Appleman]

Even in Frostbite?

Yeah, it's not really an engine limitation, but a conceptual limitation that applies to everything.

You're going to have to model and texture an order of magnitude more things if you want destruction of absolutely everything

 

[Stallion Free]

Even in Frostbite?

Did you not notice how prefab buildings were reused like crazy in the Battlefield game featuring the most destruction (BFBC2)?

 

[UrbanRats]

Yeah, it's not really an engine limitation, but a conceptual limitation that applies to everything.

You're going to have to model and texture an order of magnitude more things if you want destruction of absolutely everything

What about the DMM they had in the first Force Unleashed? Wasn't that procedural distruction?

 

[Stallion Free]

What about the DMM they had in the first Force Unleashed? Wasn't that procedural distruction?

Wasn't it limited to very specific materials? Only wood could break and metal just deformed or something like that?

And you can't procedurally generate a fridge full of food and every other detail inside household items.

 

[UrbanRats]

Wasn't it limited to very specific materials? Only wood could break and metal just deformed or something like that?

And you can't procedurally generate a fridge full of food and every other detail inside household items.

There was also jello-like stuff and glass that could break, but it wasn't the whole level like that, just very specific structures (and the framerate dipped hard whenever you were wreaking shit).

 

[shinobi602]

Added Rockstar's new engine.

 

[shinobi602]

I'm guessing the tech implemented by the guys behind "Reset" is targeted at next gen?

 

[Stallion Free]

I'm guessing the tech implemented by the guys behind "Reset" is targeted at next gen?

It's not a PC exclusive?

 

[shinobi602]

It's not a PC exclusive?

Hmm, not sure...

 

[Angelus Errare]

Everyone always forgets the best...MTFramework.

Give MTframework enough light sources and shit looks amazing. Point in case? Only every cutscene from games using it. RE5/RE6/DMC4/LP2 hell even the characters when inside the cockpit of Steel Battalion look fucking incredible

 

[Snakeyes]

Still waiting for those PC games you've been playing.

You don't need to play something exactly on par with these to be less impressed with these engines compared to current gen engines circa 2005.

 

[Zyrusticae]

Point in case?

Off-topic, but this gets on my nerves.

It's "cases in point". "Point in case" is complete nonsense, unless you were making a case and then clarifying the point afterwards (which isn't what you did).

If you're going to use idioms, please do it right!

 

[shinobi602]

You don't need to play something exactly on par with these to be less impressed with these engines compared to current gen engines circa 2005.

That makes no sense. He said PC gaming has spoiled him. Obviously he must be used to something better than what is listed in this thread to not be impressed by them and to be spoiled already.

What games does he play on PC?

 

[seattle6418]

I'm with the people asking for next gen improvements from sports games. I don't know if it's engine related - some of it is - but sports games were terrible this gen, compared to what we got in bumps of quality in previous gens.

Recently i spent a good time playing the current batch of sports games of my interest (Madden 13, NBA 2K13 and Fifa 13) and the feeling i get is that we are playing the same game since 2007.

I don't know how can people say that Fifa is a great football sim. It's a fun game, but it's terrible if you think of simulation. We still have the old floaty ball physics and the game it's programmed in such a weird pace that you end up completing a number of passes in a 10 minute game that equals what a real match has in 90 minutes.

There's no proper spacing after you past midfield, and after 20 games or so you already know all the money plays and what to do if you want to win.

It's fundamentally broken, and to me it's something that goes deeper that just AI. They need more dev time, and they seriously need a fresh start for next gen.

Madden still feels like a PS2 game. It's all about money plays, the pace of the game is also broken, meaning that scores are always exaggerated because there's too much action all the time. It's difficult to get to a grinded out game like you see in real life.

Online is a joke and due to the fact that is a football game with lots of players most of the stuff happening is automatic. My eight year old nephew who doesn't know a lot of football was playing pretty good because hey, most of the stuff on the screen the CPU was doing for him.

Basketball is my favorite sport and 2K13 is an OK game. Maybe the least worst of the bunch, but it feels dated and completely broken with money plays, patterns and glitches that make it just a game, not a simulator.

Sports games this gen have failed to read and react, to make you feel like you're in a real game. I know this is a HUGE hurdle to climb, that programming this must be way too hard, but they have to try and make improvements.

I expect next gen engines for sports games to somehow help us on this, because this generation was great for many genres, it was just a shame that for sports it was terrible.

 

[Snakeyes]

That makes no sense. He said PC gaming has spoiled him. Obviously he must be used to something better than what is listed in this thread to not be impressed by them and to be spoiled already.

It doesn't have to be something "better."

Not saying that's the case for him but I can easily understand how someone used to playing Crysis, BF3 and Witcher maxed out with mods could be less impressed with these next gen engines than a console only gamer. So in that sense, he'd be spoiled.